2

Zhipu launches GLM-5.3-FlashX with 200 tokens/s inference speed: Zhipu AI has launched GLM-5.3-FlashX, claiming inference speeds of nearly 200 tokens per second on approximately 100,000 domestic accelerators. The base Flash model is a 320-billion parameter Mixture of Experts model with 18 billion active parameters, optimized by an Infra Agent.

Zhipu AI • Pandaily

Amap launches ABot-Earth 0.7 to generate 3D cities 1,000 times faster: Amap has released ABot-Earth 0.7, a 3D-native urban world model integrated with Flying Street View 2.0. The model can generate kilometer-scale 3DGS cities on a single consumer GPU in approximately 10 minutes using satellite or text input. This process is reportedly about 1,000 times faster than traditional pipelines.

AI Models • Pandaily

Volcengine launches Doubao-Seed-2.1-pro with multimodal coding: Volcengine has fully released the Doubao-Seed-2.1-pro 0915 model featuring multimodal coding capabilities. The model demonstrated rebuilding a 280,000-line Java ERP from screen recordings and sketches, achieved an 83% mergeable rate on Luanti repo fixes, and reduced image and video inference token costs by over 30%.

ByteDance • Pandaily

PrismML compresses Alibaba's Qwen3.8 to 5.9 GB for smartphones: PrismML has released Bonsai 2 27B, a model that compresses Alibaba's Qwen3.8 27B down to 5.9 GB. This compression makes the model small enough to run on smartphones while retaining 98.2% of Qwen's original benchmark scores.

Artificial Intelligence • Techmeme

Huawei plans Q1 2027 launch of Ascend 960DT AI chip to rival Nvidia: Huawei is accelerating the launch of its next-generation Ascend 960DT AI chip, targeting a release in the first quarter of 2027. The move is part of the company's push to compete with Nvidia and narrow China's AI computing gap with the United States.

Hardware • TechCrunch AI

Huawei launches Ascend 960 SuperPoD with NPO technology for AI scaling: At the HUAWEI CONNECT 2026 event, Huawei Deputy Chairman Ken Hu announced the Ascend 960 SuperPoD. The new system utilizes Near-Pool Optoelectronics (NPO) technology to address the computing power demands of scaling artificial intelligence models.

Hardware • TechNode

DeepCybo open-sources PhysBrain 1.5 physical foundation model: DeepCybo has open-sourced PhysBrain 1.5, an embodied physical foundation model available in 2B and 8B parameter sizes. The model achieved an average score of 72.5 across 28 public embodied benchmarks and features unified understanding, action, and future-state tokens alongside an evaluation kit.

DeepCybo • Pandaily

Light Origins open-sources LightNav-0 generalist navigation model: Light Origins has open-sourced LightNav-0, a generalist navigation model based on Qwen3-VL-4B. Trained via Real2Sim2Real on over 2,000 scenes and 4,000 hours of VLA data, the model leads 10 monocular navigation benchmarks and supports zero-shot body transfer.

AI Models • Pandaily

Researchers build Pareto atlas to optimize LLM inference configurations: Researchers have developed a cost, quality, and latency Pareto atlas to identify optimal LLM inference configurations under various deployment constraints. The team measured 54 configurations of Qwen2.5-7B-Instruct on vLLM across L4, A100, and H100 GPUs to calibrate a simulator.

Research • arXiv

Shanghai AI Lab open-sources 744B parameter Atria Dawn Preview model: Shanghai Artificial Intelligence Laboratory has open-sourced Atria Dawn Preview under the MIT license. The 744-billion parameter Mixture-of-Experts agentic model is built on GLM-5.2, featuring a 256K context window, FP8 weights, and an AutomationBench score of 53.8.

Artificial Intelligence • Pandaily